An effective approach for providing diverse and serendipitous recommendations

Abstract

Over the years, recommendation systems successfully suggest relevant items to its users using one of its popular methods of collaborative filtering. But, the current state of recommender system fails to suggest relevant items that are unknown (novel) and surprising (serendipitous) for its users. Therefore, we proposed a new approach that takes as input the positive ratings of the user, positive ratings of the similar users and negative ratings of the dissimilar users to construct a hybrid system capable of providing all possible information about its users. The major contribution of this paper is to diversify the suggestions of items, a user is provided with. The result obtained shows that as compared to general collaborative filtering, our algorithm achieves better catalogue coverage. The novelty and serendipity results also proved the success of the proposed algorithm.